Consider this. You are about to buy a new car and have finally made up your mind which make and model you are going to buy. You then set off for work the next morning, and every second car you see on the road is the same model as the one you have chosen.
Why is this? Has the model gained in popularity overnight? Most likely not. In fact, your eyes are seeing what you have unconsciously programmed your brain to see. In other words, you have fallen victim to confirmation bias.
So what is confirmation bias?
Confirmation bias happens when you have already decided on an answer; you will then subconsciously try to find the data that confirms that answer, and will disregard any data to the contrary. Needless to say, that is not quite the right approach to working with data.
It is natural to have a gut feel about things; most of us do. But the key to avoiding confirmation bias is to be flexible with our gut feeling, to allow for contradictions. Fix your decision criteria first and set that in stone. Only consider challenging your gut feeling if the information that you extract from the data informs you to do so. The goalpost should not move once set, so once it is, this is when you should delve into data analysis to come to a decision.
Are you skewing your decisions?
To make an informed decision we need good data. Good data needs to be accurate, complete, there needs to be enough of it, and it should also be representative. What that last term means in this context is that the sample we are analysing should represent the population as a whole, otherwise the outcome will be skewed.
If we are to assess the popularity of cricket as a sport, and we conduct a survey outside a cricket club, it is clear that the results will over-inflate cricket’s popularity. That is because we are only asking people who are, in all likelihood, interested in the sport already. In other words, our assessment will suffer from selection bias. The correct selection process would be to ensure that all different groups present in a population are represented in the sample that will be used for analysis. That will be a fair sample, and it will give us a fair outcome.
What if the decision outcome does not go as expected?
Outcomes are important. If you consider our actions at work and at home, we look towards achieving a set of outcomes. After all, an outcome is what we are working towards and hoping to influence in a positive way via the decisions we are making. But what if the expected outcome does not materialise, and instead something quite the opposite happens?
If we are to assess the popularity of cricket as a sport, and we conduct a survey outside a cricket club, it is clear that the results will over-inflate cricket’s popularity
Let’s say every month you decide to put your excess income into your savings account, while your neighbour decides to buy lottery tickets instead. Now if your neighbour does go on to win the jackpot, does this mean your original decision was wrong? Should you change your mind based on this outcome, stop saving and buy lottery tickets from now on? Obviously not, as that would be an extreme case of outcome bias messing with your perfectly rational decision.
In other words, a decision should be judged on the merit of information available at the time of making it. It would be wrong to judge, and often nothing short of disastrous to change a decision, on the basis of a random outcome.
Introducing the family of decision biases
The cases we have discussed so far are all instances of decision bias. And these are not the only types of decision bias. There are several others, including but not limited to:
- Survivorship bias: a type of statistical bias in which you concentrate on the part of the data set that was already preselected in some way, and disregard the part of the data set that was not (because it is no longer visible). This can lead to over-optimism as failures are ignored – for example, excluding businesses that no longer exist from an analysis of financial performance in a certain sector.
- Recall bias: this is a commonly occurring error in a research or interview scenario, when a person’s possibly inaccurate recollection of a past event is relied upon as evidence.
- Funding bias: the tendency for a study or research exercise to support the interests of the study’s financial sponsor.
- The Ikea effect: a cognitive bias in which consumers place a disproportionately high value on products they partially created, such as in the case of Ikea’s flat-pack furniture. A 2011 study found that respondents were willing to pay 63% more for furniture they had assembled themselves than for equivalent pre-assembled items.
As decision-making is inherently a cognitive activity, individual subjective predispositions can disrupt objective judgement. This in turn could result in misrepresentation of data, incorrect decisions, and everything else that follows. It is evident that it is quite easy for these decision biases to creep in, whether in day-to-day life or when handling data and looking for outcomes at work.
To avoid decision biases, it’s imperative to be aware of them and to know in what way they can influence your decision-making. It will then be a case of taking the necessary actions to eliminate them, via steps like data source verification and random sampling, exploratory rather than confirmatory data analysis, reviewing findings with peers and involving all stakeholders in decision-making to gain different perspectives, in order to arrive at a fair and robust outcome.